Impact of Spatial Frequency Based Constraints on Adversarial Robustness
This work addresses adversarial robustness in machine learning models by exploring frequency-based constraints, but it is incremental as it builds on known cognitive science insights without introducing a new paradigm.
The paper investigates how training models to use specific spatial frequency ranges affects adversarial robustness, finding that the impact varies significantly across datasets, with differences in adversarial accuracy up to 0.41.
Adversarial examples mainly exploit changes to input pixels to which humans are not sensitive to, and arise from the fact that models make decisions based on uninterpretable features. Interestingly, cognitive science reports that the process of interpretability for human classification decision relies predominantly on low spatial frequency components. In this paper, we investigate the robustness to adversarial perturbations of models enforced during training to leverage information corresponding to different spatial frequency ranges. We show that it is tightly linked to the spatial frequency characteristics of the data at stake. Indeed, depending on the data set, the same constraint may results in very different level of robustness (up to 0.41 adversarial accuracy difference). To explain this phenomenon, we conduct several experiments to enlighten influential factors such as the level of sensitivity to high frequencies, and the transferability of adversarial perturbations between original and low-pass filtered inputs.